Martin Wunderlich's Blog

By now it probably has become common knowledge that you can use google translate to entertain yourself with some beatbox sounds. The google translate blog points to a few other very creative uses of the tool, such as ordering food in Hindi, singing a Taiwanese song, understand what your pet has been trying to tell you.

Cheers,

Martin

The website heise.de, run by a German publishing house, is one of the most important sources of information for all things IT in Germany. They are also the makers behind the magazine “C’T” – a near must-read for German IT professionals. I am mentioning this here, because I have recently noticed that they have published a good article on open-source software for translators:
http://www.heise.de/open/artikel/Open-Source-fuer-Uebersetzer-1204029.html
http://heise.de/-1204029

One of the links on the second page of the article points to my old link collection of open-source software, tools, and utilities for translators (also in German: DE). There is a reason why the site with the link collection isn’t active anymore and has been replaced by this blog: I haven’t had the time to properly maintain the list for the past few years. A lot has happened since it was last updated. However, quite a few of the links are still valid and the reader might discover something valuable. So, have a look around.

(And whenever I find the time, I will convert the stuff to a Wiki so that the maintenance work doesn’t rest on my shoulders alone   )

M

Minor annoyances, when consciously observed and acted on, can lead to great positive change.

Checking various email addresses in a number of environments (iPhone, private Macbook, work Macbook, webmail), I found myself annoyed by the following: emails that I had read on one device would still be marked as unread on other devices. I had heard of the IMAP standard (as opposed to POP/SMTP), whereby the clients are kept in synch with the server instead of downloading emails. Re-configuring the email providers and email clients was quick enough.

However, I ran into a problem with the iPhone: For some reason, when entering the account details, the iPhone will auto-select the account to be a POP account. Now here is the trick: This auto-selection only works, if you enter the _correct_ account details. So, in order to self-select IMAP over POP, I misspelled the password and – tada – the iPhone presented me with the full config details of the account and I was able to select IMAP.
Just in case, anyone else runs into the same problem…

Cheers,

Martin

As per this here blog post:
http://blog.gts-translation.com/2011/04/01/apple-buys-sdl-to-create-new-language-division/

In a similar move, Welocalize has also bought SDL, in addition to buying Moravia:
http://www.linkedin.com/groupAnswers?viewQuestionAndAnswers=&discussionID=49026100&gid=44105&commentID=35663163&trk=view_disc

What a coincidence….

There are probably quite a few translation and localization managers out there, who often have to fend off that “idea” to use machine translation. Colleagues or managers look at the likes of Google Translate and they look at the time and money spent on creating human-made translations – and suddenly machine translation seems like a fantastic option.

Translation guy has a few insights on this in his recent blog post.  As usual, the post is well written, insightful and does not shy away from critique. There are many other examples of machine translation gone wrong, such as the infamous “translate server error“.

In essence, there are four use cases at present where machine translation does add value:

- so-called “gisting”, i.e. if you want to quickly and roughly understand what a text in a certain language says;

- translation of highly controlled and structured source content (e.g. weather reports or virus warnings); cf. “controlled language“.

- pseudo-localization, i.e. testing a piece of software with regards to internationlisation or localisability.

- translation of content that wouldn’t otherwise get translated: this is a grey one, though. It certainly doesn’t apply to highly visible or printed content. You might want to use MT on support cases or forum posts, but they should be clearly labeled as having been translated using MT.

If someone thinks that you should use MT for anything else, simply provide them with a few examples, back-translations etc. This will quickly change their minds. Or refer them to the comment to this post which is a back-translation from this post translated into German and then English again using google translate.

M

Alain de Botton is a Swiss-born philosopher and writer, residing in the UK, who has the great gift of putting new or forgotten thoughts into simple words – words that are as simple as possible, but not simpler.
Nowadays, where we have many professors of philosophy and few philosophers, de Botton has stepped up to the task of analysing, commenting on and evaluating daily life, be it by writing in an airport or by re-reading Proust. His writing often takes the form of a mock-self-help book, but avoiding the usual banalities and superficialities of that genre and instead presenting a critical view on Western life-styles and on their inherent lack of life and style.

This here is a great speech on why pessimism (or at least stoicism) is an appropriate attitude and will prevent unhappiness:

A fantastic speech to listen to when things don’t go as you believe they should, be it in love life or at work.

M

(in response to this here blog post on Content Rules)

Industry associations provide more than networking opportunities and conferences. LISA had it in their name that they were primarly a standards setting organisation and we definitely do need standards bodies. LISA’s demise might have something to do with the organisation not focusing on this core activity that much anymore in recent years. They did provide us with TMX, but TMX exists in many different levels and versions and hasn’t fully delivered on the promise of lossless TM exchange. How about GMX with its various flavours? How many tools are actually using it? How about TBX (which is way to complicated for daily use)?
LISA could have concentrated much more on making the existing standards rock-solid. In this regard, LISA should have been looking to the Unicode consortium as a model, for instance.

LISA have also missed the way into the cloud. We have a number of web-based TM technologies competing today (TDA, Wordfast, Globalsight etc.), but no standards for the web services APIs that are used to access these systems and applications. In a similar manner, we will probably soon see web-based terminology databases (and Wiktionary might provide the foundation for these). Again, this is an area where LISA as an independent organisation could have been leading the way by defining API standards and working with tools developers to implement them.

It is a pity to see LISA disappear and I do hope that another organisation will continue to maintain and develop the exchange standards. Because we need them badly, especially in an interconnected localization world that is more and more moving into the cloud.

Cheers,

Martin

I have just read a blog post on Mashable with the intriguing title “10 Ways to Turn Your Local Business Into a Global Success“. The author mentions several good points, some of which are not immediately obvious. For instance, allowing for Bidi languages, such as Arabic and Hebrew, in your web layout or respecting local preferences for payment systems.

However, the first item on the list I have to disagree with. The author advises to quickly translate a business’s website on the cheap, using services like myGengo or machine translation. With more than 10 years localization experience under by belt and the last three years dedicated to web localization, I can tell for sure that this is a recipe for disaster. Marketing-oriented web copy does not lend itself easily to MT. MT is usable for translating highly controlled source text or for the purpose of gisting, but not much more. Besides, there is a lot more involved than just translating the text when localizing web content. There are editing and review steps, web production, and other technical work.

Web localization is a complex craft that should not be diminished by suggesting that it could be achieved quickly and cheaply in the manners described by the author, while achieving results that provide professional quality. Maybe the blog post really is about only “9 Ways to turn your local business into a global success” and “One sure-fire way to turn your local business into a global failure.” IMHO.

Cheers,

Martin

One of the main advantages of our Globalsight installation is the ability to pull custom reports from the database by means of a simple SQL query. This one here, for instance, will list all currently available activities and the users that they are available to by target locale:

SELECT job.id, job.name As 'Job name', CONCAT(locale.iso_lang_code,'-',
locale.iso_country_code)
As 'Target Locale',
jbpm_taskinstance.name_, jbpm_pooledactor.actorid_ As 'Available to'
FROM job, workflow, task_info, locale, jbpm_taskinstance, jbpm_task,
jbpm_pooledactor, jbpm_taskactorpool
WHERE jbpm_pooledactor.id_ = jbpm_taskactorpool.pooledactor_ AND
jbpm_taskactorpool.taskinstance_ = jbpm_taskinstance.id_ AND
jbpm_taskinstance.actorid_ IS NULL
AND workflow.job_id = job.id AND
workflow.iflow_instance_id = task_info.workflow_id AND
jbpm_taskinstance.task_ = jbpm_task.ID_ AND
jbpm_task.tasknode_ = task_info.task_id AND
workflow.target_locale_id = locale.ID AND
job.state = 'DISPATCHED'

There are many other examples listed in the help wiki: 
http://globalsight.com/wiki/index.php/Useful_SQL_database_queries

This only works, because we have our own instance.
If you are running GS as a multi-tenant service, it won't work. But if you have your own installation,
this works fine and allows for very powerful reporting.
(I find this ability so cool that I am even writing a blog post about it late on a Friday night.)

Cheers, 

Martin

We have been using Globalsight as our central localisation management system in marketing for about 5 months now. The migration was completed during a two month project in summer 2010. We were looking for a system that would provide more flexibility in terms of process modeling and would have an open architecture that we can build a proper localisation platform on.

Note that this blog post is not meant as a commercial advertisement for the system. I just want to give other enterprise localisation managers in the IT industry some ideas about what was involved and what to watch out for.

The scope was something like this (as determined retrospectively):

• SaaS- based deployment of 5 servers in fully managed data center (prod and dev/test system)
• Creation of 44 localization profiles
• Building 200 custom workflows
• 49 unique file processing rules
• Setup and training for 55 users distributed around the globe
• Migration of over 600 MB of translation memory data
• Terminology database population with over 8000 terms

So, what were the good, the bad and the ugly of the migration project?

• On the plus side, I created only a rough project schedule at the start in collaboration the different people involved (system vendor, localisation vendor, procurement, web dev, IT security…), so that we would maintain the right level of flexibility and agility. We completed the project nearly on time after 2 months (ok, it was a few days late, but that’s not bad for a project with this level of complexity).
• When the system went live, all the assets and most of the required workflows were set up, all the internal users in salesforce and all the vendor PMs were trained and ready to use the system.
• For the purpose of training internal reviewers, I created a video that they could watch at their convenience (important when the reviewers are distributed around the globe).
• We also ran a successful configuration workshop in the form of a 1-day face-to-face meeting. This helped us to work out the confg requirements quickly and immediately start implementing them in the system.
• Most importantly: We had a great project team and everyone was highly dedicated and committed to the project. All the required skills and departments were represented and the determination of the team members to get things done blew me away many times.
• Regular meetings or status calls were running smoothly, on a weekly basis initially and more frequently towards the end of the project.
• We were able to complete the project in a globally distributed team effort, without a so-called “war room” and without people spending many late nights or working weekends.

What could we have done better? Captain Hindsight would say:

• Allow sufficient time for testing, of both the dev/test and of the actual production system. If the overall migration takes two months, at least two weeks should be allowed for thoroughly testing the system. Make sure you have a test plan worked out and test tasks clearly assigned. Also, make sure to agree timelines with the testers for completing these tasks. The time that we had planned for testing was a bit too short.
• Don’t train the users too soon. We trained the users several weeks before the system went into production. By the time they started working with the system, we had to retrain some of them (which wasn’t that hard, due to the video-based training, but nevertheless an inconvenience).
• Don’t train the users too late. I had assumed that the translators and production teams at the vendor side could start using the system shortly after go-live. As it turned out, they weren’t ready quite as quickly, because they needed more time than expected to adjust to the new way of working.
• If you have a hard deadline, make sure you have sufficient buffer time in the project plan to allow for the unforeseen to happen, especially when you are working in a complex corporate environment. In our case, we lost a few weeks due to some unexpected internal procedures (which I can’t disclose here).
• Don’t expect things to work perfectly fine after go-live, even with enough testing. There will be a period of making minor incremental adjustments to the system, based on what you learn during the production use. There might locale pairs missing or workflows need to be adjusted slightly. This period will most likely take a couple of weeks.

Overall, I would say the Globalsight migration project was a great success and hopefully others who are planning a similar move will be able to benefit from the experience as described here.

Cheers,

Martin